thinhpt/merged_12052026

该数据集是一个机器人控制数据集，由LeRobot项目创建，专注于机器人模仿学习或强化学习任务。数据集包含300个回合，总计111273帧，数据以parquet格式存储，视频以mp4格式存储，帧率为30fps。机器人类型为so101_follower，特征包括：动作（action）为6维浮点数组，表示机器人关节位置（肩部平移、肩部升降、肘部弯曲、腕部弯曲、腕部旋转、夹爪位置）；状态观测（observation.state）同样为6维浮点数组，表示机器人关节位置；多视角图像观测，包括顶部摄像头（observation.images.top）、腕部摄像头（observation.images.wrist）和侧面摄像头（observation.images.side），均为480x640分辨率、3通道的彩色视频；以及时间戳、帧索引、回合索引、任务索引等元数据。数据集适用于机器人控制、视觉感知和时序决策等研究。This dataset is a robotics control dataset created by the LeRobot project, focusing on robot imitation learning or reinforcement learning tasks. It contains 300 episodes, with a total of 111273 frames, stored in parquet format for data and mp4 format for videos, at a frame rate of 30fps. The robot type is so101_follower, and features include: action as a 6-dimensional float32 array representing robot joint positions (shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, gripper); observation.state as a 6-dimensional float32 array also representing joint positions; multi-view image observations from top camera (observation.images.top), wrist camera (observation.images.wrist), and side camera (observation.images.side), all with 480x640 resolution and 3-channel color video; and metadata such as timestamp, frame index, episode index, and task index. The dataset is suitable for research in robot control, visual perception, and temporal decision-making.